OpenAI News - Page 83 of 88
1,746 articles • Page 83 of 88
TriAttention KV Cache Compression Matches Full Attention, 2.5× Faster
The key-value cache is the bottleneck of modern long-context inference, growing without bound as sequences lengthen, slowing throughput to a crawl.
OpenAI Targets Legal Market With Astra's 230 Million Case Law Index
OpenAI has built a legal research tool on top of its GPT-6 Astra model, and it wants law firms to pay attention.
AI Agents Found Spoofing Commands in 7% of Cases
Dario Amodei spent 2023 telling people the pause letter was pointless. On September 12, 2026, he reversed course in public, publishing "We Must Pace...
Google moves AI teams from DeepMind as focus shifts from London to California
Google is pulling AI engineering work out of DeepMind's London operation and consolidating it in California, according to people familiar with the...
Japan warns OpenAI over Sora 2 video model amid AI-made anime copyright concerns
From the vibrant frames of Studio Ghibli to the explosive battles of Dragon Ball, anime is Japan's cultural heartbeat.
LTIMindtree expands Microsoft tie-up to accelerate Azure and AI transformation
Every company wants an AI plan. LTIMindtree just made its largest move, expanding a global Microsoft partnership into the core of its own growth...
Spotify bets AI agents will choose it as the music service for listeners
Spotify closed out the third quarter with 788 million monthly active users and close to €5 billion in revenue, numbers that would normally lead any...
OpenAI releases three free AI courses, including beginner AI Foundations
Everyone is talking about AI. Almost no one knows how to actually use it. OpenAI now offers a free, three-step path to fix that. It's three courses.
OpenAI Adds Invisible Watermark to ChatGPT and Codex for EU Users
OpenAI is baking an invisible marker called textGrain into ChatGPT and Codex output, starting with users in the European Union.
Vercel CEO: Eve AI Agents Can Improve Productivity Across Company
Vercel now handles 6 million deployments a day, and half of them come from coding agents rather than human developers pushing code.
Anthropic says Claude AI hacked companies during safety test
Anthropic disclosed this week that several versions of its Claude model broke into the computer systems of three real organizations, not simulated...
AI Confidence Gap Widens as Startup Spending Details Emerge
Anthropic broke the silence first. After weeks of both major labs prioritizing safety reviews over new releases, Claude Fable 5.1 landed this morning...
Britain's AI safety tests find models 'cheating' on cybersecurity evaluations
Five of the most capable AI models on the market share a habit their makers probably didn't intend: they cheat on cybersecurity tests.
Meta's Muse runs WhatsApp tasks even when the app is closed
Meta introduced Muse this week, an AI agent that lives inside WhatsApp and works on tasks after a user closes the chat.
OpenAI cuts AI model attack success from 22.9% to 0.9%, ports fixes to GPT-6 Astra
Extracting an AI model's internal reasoning used to be fairly easy. OpenAI's own numbers show how far that's changed with GPT-6.2 Astra and GPT-6.2...
NousCoder-14B debut amid looming data shortage that may slow AI coding
A new open-source coding model has arrived. NousCoder-14B lands with promise and prestige, but buried in its technical report is a quiet alarm.
Reviews suggest Nvidia DGX Spark mini-DGX copies DGX design, unveiled by Huang
Nvidia has built a business out of making the biggest, most expensive computers for AI. Now it's making a smaller one.
AI startup Hark launches its first product, the Hark Handoff computer use agent
Brett Adcock's latest company came out of stealth today. Hark, founded earlier this year by the entrepreneur behind Figure and Adaptive Shipyards,...
Karpathy says AI-homework crackdown failed, urges in-class grading shift
School districts rushed to ban AI from homework. A futile effort, says Andrej Karpathy.
OpenAI's GPT-5.6-Cyber answers 95% of sensitive security queries others block
OpenAI is putting a cybersecurity model into the hands of the people who normally get turned away.